Wavelet Analysis Combined with Artificial Neural Network for Predicting Protein–Protein Interactions

Abstract

In order to solve the prediction problem of interaction between proteins, we use a wavelet coefficient combined with artificial neural network method, improving the prediction accuracy of the problem of protein–protein interactions. By introducing the Biorthogonal Wavelet 3.3 coefficients as the feature extraction method and the three-layer feedforward neural network as a classifier, we solve the problem of protein interaction effectively. Using the Human dataset verifies the validity of this method. Through testing the Human dataset, using Biorthogonal Wavelet 3.3 coefficient combined with the three-layer feedforward neural network, solve the prediction problem of protein interactions with well results. This combination of wavelet coefficients and the three-layer feedforward neural network to predict protein interaction problem is an effective method. At the same time, compared with other prediction methods, this method performs at least 4 % higher accuracy than the better accuracy of auto-covariance (11) combined with PNN on the same dataset.